Muddler is a tool for sharing derived data.
It is sometimes necessary to share data publicly that has been derived from a non-public source without sharing the source itself. Muddler solves this issue by "subtracting" the source data from the target (or derived) data, a process we call muddling. This generates a muddled package that can be distributed publicly. In order to retrieve the target data, users must first acquire the derived data through proper channels and can then unmuddle the muddled package through Muddler to generate the target data.
This is particularly useful when the derived data cannot be trivially reconstructd from the source data such as human annotated/modified data.
Please make sure that the source license permits creating derived work!
Muddler requires Python 3.6+ to install.
To install using pip, just run:
pip install muddlerUsage: muddler muddle -s <SRC_PATH> -t <TRG_PATH> <MUDDLED_PATH>
muddler muddle -c <CONFIG> -s <SRC_PATH> -t <TRG_PATH> <MUDDLED_PATH>
muddler unmuddle -s <SRC_FILE> -m <MUDDLED_PATH> <TARGET_OUT>
muddler (-h | --help)
muddler (-v | --version)
Options:
-h, --help
Print help message.
-v, --version
Print muddler version
-c <CONFIG>
Path to muddler config file.
-s <SRC_PATH>
Path to source file or directory. When <CONFIG> is not specified in
muddle mode, SRC_PATH must point to a file and not a directory.
-t <TRG_PATH>
Path to target file or directory. When <CONFIG> is not specified in
muddle mode, TRG_PATH must point to a file and not a directory.
-m <MUDDLED_PATH>
Path to muddled package to be unmuddled.
Muddler runs two modes: muddle mode for generating muddled packages, and unmuddle mode to extract targets from a muddled file.
The simplest example for muddling is when both the source and the target are single files.
For example:
muddler muddle -s /path/to/source_file -t /path/to/target_file /path/to/my_package.muddleNote that the muddled package doesn't have to end with the .muddle extension.
When at least one of either the source or target is a directory (ie the source data is composed of multiple files), we must additionally pass a config file to muddler.
For example:
muddler muddle -c /path/to/config_file -s /path/to/source_dir -t /path/to/target_file /path/to/my_package.muddleThe config file provides muddler with a mapping on which source file(s) where used to derive the target(s). See the Config Format section for more information.
To unmuddle a muddled package, one must first acquire the source files from which the muddled data is derived from. The acquired source file or directory must be exactly the same both in directory structure, file names, and file contents (byte for byte).
Once the source files have been acquired, the muddled package can be unmuddled by running:
muddler unmuddle -s /path/to/source -m /path/to/my_package.muddle /path/to/target_outputThe generated target will either be a single file or a directory depending on the target used for muddling.
Below is a documented configuration file that structure in general:
- This is a comment! All comments begin with '-' must be on their own individual lines.
- Comments are ignored by muddler and are used for documentation and organizational purposes.
- Comments can be preceeded by whitespace as well.
- The first section in a config file is a header that tells muddler what kind of source and
- target to expect as well as which muddling algorithm to use.
- The TARGET_TYPE and SOURCE_TYPE headers tell muddler whether to expect a file or directory
- for target and source respectively. A value of 'file' indicates a single file, while 'dir'
- indicates a directory.
##TARGET_TYPE dir
##SOURCE_TYPE dir
- The algorithm version tells muddler what algorithm to use to create the muddled package.
- At the moment only one algorithm ('1') is available but the field is required for backwards
- compatibility when new algorithms are added.
##ALGORITHM_VERSION 1
- After specifying the header, we can start specifying targets and their respective sources.
- For a 'dir' target, each target entry consists of a relative path within the target directory.
- For example, if the absolut path to the target directory is '/home/username/target', then
- the below target entry specifies source for the file at '/home/username/target/target_01.txt'.
- Note that each target path MUST start with '/'.
- All speaces after #TARGET and the first '/' character are ignored, but there needs
- to be at least one.
- Note that any whitespace at the end of a line is NOT IGNORED.
- This is because valid filenames can include whitespace.
#TARGET /target_01.txt
- Each target must provide a list of sources used to derive it.
- This list has to be unique.
- Similar to targets, each source is a relative path to a file within the source directory.
- They must also start with a '/' character and proceeding whitespaces are NOT IGNORED.
- Preceeding whitespace is ignored though.
/source_01.txt
/source_02.txt
- Both target and source paths can point to files in subdirectories.
#TARGET /sub/target_02.txt
/source_01.txt
/sub/source_03.txt
- Each target entry must have a unique target path.
- So adding the following entry would cause an error:
#TARGET /target_01.txt
/source_01.txt
/source_02.txt
Additional rules apply when either target or source are single files.
When target is a single file, the config file should include only one target entry where the target path is '/' as so:
##TARGET_TYPE file
##SOURCE_TYPE dir
##ALGORITHM_VERSION 1
#TARGET /
/source_01.txt
/source_02.txt
If source is a single file then target entries should not be followed by any source lines as so:
##TARGET_TYPE dir
##SOURCE_TYPE file
##ALGORITHM_VERSION 1
#TARGET /target_01.txt
#TARGET /target_02.txt
#TARGET /sub/target_03.txt
Muddler is available under the MIT license. See the LICENSE file for more info.